An understanding of cellular signalling from a systems-based approach has to be robust to assess the effects of point mutations in component proteins. Outcomes of these perturbations should be predictable in terms of downstream response, otherwise a holistic interpretation of biological processes or disease states cannot be obtained. Two single, proximal point mutations (S252W and P253R) in the extracellular region of FGFR2 (fibroblast growth factor receptor 2) prolong growth factor engagement resulting in dramatically different intracellular phenotypes. Following ligand stimulation, the wild-type receptor undergoes rapid endocytosis into lysosomes, whereas SWFGFR2 (the S252W FGFR2 point mutation) and PRFGFR2 (the P253R FGFR2 point mutation) remain on the cell membrane for an extended period of time, modifying protein recruitment and elevating downstream ERK (extracellular-signal-regulated kinase) phosphorylation. FLIM (fluorescent lifetime imaging microscopy) reveals that direct interaction of FRS2 (FGFR substrate 2) with wild-type receptor occurs primarily at the vesicular membrane, whereas the interaction with the P253R receptor occurs exclusively at the plasma membrane. These observations suggest that the altered FRS2 recruitment by the mutant receptors results in an abnormal cellular signalling mechanism. In the present study these profound intracellular phenotypes resulting from extracellular receptor modification reveal a new level of complexity which will challenge a systems biology interpretation.

INTRODUCTION

FGFRs [FGF (fibroblast growth factor) receptors] are receptor TKs (tyrosine kinases) involved in proliferation, growth inhibition and differentiation in different cell types [1]. All members of the FGFR family have an extracellular ligand-binding region, a transmembrane domain and an intracellular TK domain. The extracellular portion of the FGFRs [which comprises three Ig (immunoglobulin)-like domains] interacts with FGF ligands. So far 23 distinct FGFs have been identified in a variety of organisms [2,3]. Each FGFR isoform has restricted tissue-specific expression and ligand-binding properties [1]. The growth factors alone are poor ligands for FGFRs without the cell surface accessory HSPG (heparan sulfate proteoglycan) molecules [1,4,5]. The crystal structures of receptors bound to growth factors suggest that HSPGs form part of a ternary complex performing a dual function as both an augmenter of the FGF–FGFR interaction and a promoter of FGF–FGFR dimerization [6,7]. Receptor dimerization leads to autophosphorylation of seven conserved intracellular tyrosine residues [8] which serve as recruitment sites for SH2 domain-containing proteins to initiate downstream signalling [9]. FRS2 (FGFR substrate 2) has been shown to be constitutively bound to FGFR1 [10]. On phosphorylation, FRS2 serves as a docking site for a number of intracellular proteins including Grb2 (growth-factor-receptor-bound protein 2) [11], SHP-2 (Src homology 2 domain-containing protein tyrosine phosphatase 2) [12,13] and c-Cbl (casitas B-lineage lymphoma) [14]. The recruitment of the Ras activator SOS (Son of Sevenless) through a Grb2–FRS2 complex is thought to be the primary route for activation of the MAPK (mitogen-activated protein kinase) pathway.

The present study examines the effects of perturbing the growth factor binding, particularly with respect to the longevity of interaction. To assess how changing the temporal parameters of receptor signalling affects the downstream intracellular response, we made two point mutations (S252W and P253R) in the linker region between the second and third Ig-like domains of the extracellular portion of FGFR2. These mutations had previously been shown to give rise to higher-affinity binding derived primarily from extended off rates for some FGF ligands, resulting in prolonged receptor engagement [1517] and result in the phenotypically well characterized Apert's syndrome [1,1821]. X-ray crystallographic detail on the complexes of wild-type and mutants of FGFR2 with FGF reveal that there are no gross conformational changes of the interacting proteins on complex formation, and the mutations impose limited changes to the receptor–growth-factor interface [1517]. Thus this receptor system provides an ideal model to probe the question of how a single point mutation which alters the temporal engagement of the growth factor affects protein interactions in intracellular signalling [22].

Using stable HEK-293T cells [human embryonic kidney-293 cells expressing the large T-antigen of SV40 (simian virus 40)] transfected with wild-type, S252W or P253R FGFR2 as our model, we investigated differences in the intracellular response to FGF binding. Comparison of the wild-type and mutant receptors reveals strikingly distinct temporal and spatial distribution of the ligand-activated receptor and the receptor-associated tyrosine-phosphorylated proteins. Perhaps more surprising is the observation of idiosyncratic behaviour resulting from the individual mutants. The two mutant receptors appear to show altered glycosylation patterns and ligand-stimulated phosphorylation, as well as markedly up-regulated downstream signalling compared with the wild-type FGFR2. In addition, FRS2 shows clear differences in localization patterns and mode of interaction with the three receptors. The present study emphasizes the complex intracellular outcome resulting from limited modification to the extracellular domain and alteration of the time of engagement of the growth factor. This example suggests that, since a dramatic perturbation of intracellular signal transduction can arise from such a modest change to the receptor, the ultimate description of signalling from a systems biology perspective will be extremely challenging.

EXPERIMENTAL

Materials

A mouse monoclonal antibody directed against phosphorylated ERK (extracellular-signal-regulated kinase) and a rabbit polyclonal antibody against total ERK were purchased from Cell Signaling Technology. A goat polyclonal antibody against GFP (green fluorescent protein) was purchased from Rockland Immunochemicals. Anti-phosphotyrosine (pY99), anti-FGFR2 (Bek), anti-Grb2 (rabbit polyclonal), anti-FRS2 (rabbit polyclonal) and anti-tubulin antibodies were purchased from Santa Cruz Biotechnology. A mouse monoclonal anti-FRS2 antibody was also purchased from R&D Systems. The pY99 anti-phosphotyrosine antibody was labelled with the sulfonamide dye Cy3 (Amersham Pharmacia) as described previously [23]. For the labelling of specific molecules, a Cy3 antibody was used (molar ratio of 20:1). Tunicamycin was purchased from Calbiochem. LysoTracker Red DND99 was purchased from Invitrogen. GST (glutathione transferase)-p13suc1 was purchased from Upstate.

Cloning

The full-length FGFR2 (IIIc) and S252W cDNA were a gift from John Heath (Department of Biochemistry, University of Birmingham, Birmingham, U.K.) and was PCR-amplified using flanking primers with BamHI and HindIII restriction sites. The amplified DNA was inserted in-frame into the BamHI/HindIII site of the pEGFP-N2 vector (Clontech). The P253R mutation was created using a site-directed mutagenesis kit (Stratagene) and was confirmed by DNA sequencing. The WTΔC (similar to the K-sam-IIC3) was generated using an internal restriction site within the FGFR2 cDNA. The wild-type FGFR2 in pEGFP was digested with EcoRI/BamHI to remove a 180 bp DNA fragment from the C-terminus of the receptor. The monomeric red fluorescent protein (termed mRFP-C) in pcDNA3.1(+) was provided by Tony Ng (Randall Division of Cell and Molecular Biophysics, King's College London, U.K.) [24]. The full-length human FRS2 cDNA [GenBank® accession number: BC021562 (IMAGE: 4556225)] was obtained from HGMP (Human Genome Mapping Project, Hinxton Hall, Cambridge, U.K.), PCR-amplified using flanking primers and subcloned in-frame into the mRFP-C vector.

Cell Culture

HEK-293T, HeLa and 3T3-L1 fibroblast cells were maintained in DMEM (Dulbecco's modified Eagle's medium) supplemented with 10% (v/v) FBS (foetal bovine serum) and 1% antibiotic/antimycotic (Cambrex) in a humidified incubator with 5% CO2. ROS17/2.8 osteosarcoma cells were grown in Phenol-Red-free DMEM supplemented with 10% (v/v) FBS and 1% antibiotic/antimycotic (Cambrex). HEK-293T cells were transfected with DNA encoding the WTFGFR2–GFP (wild-type FGFR2 fused to GFP), S252WFGFR2–GFP or P253RFGFR2–GFP using Lipofectamine™ 2000 (Invitrogen) according to the manufacturer's instructions. Stable cells were generated by G418 selection (800 μg/ml) for 2 weeks with a change of medium every 4 days. The resulting antibiotic-resistant cells were plated in 96 well plates in limiting dilution. Individual wells with various degrees of fluorescence were selected and expanded. Cells with equivalent levels of receptor were selected for comparison experiments. FRS2–RFP was transiently transfected into HEK-293T cells overexpressing WTFGFR2–GFP, S252WFGFR2–GFP or P253RFGFR2–GFP respectively using Lipofectamine™ 2000 as before. At 24 h post-transfection, cells were detached and seeded on to glass coverslips and allowed to grow for a further 48 h before serum starvation and stimulation.

Cell lysis, immunoblotting and pulldown assays

Cells were grown in 10 cm dishes, serum-starved overnight and then stimulated by the addition of 10 ng/ml FGF2 or FGF9 (Peprotech or R & D Systems respectively), and were incubated at 37 °C for the indicated time period. For tunicamycin treatment, cells were seeded and allowed to attach overnight, incubated with 3 or 10 μg/ml tunicamycin for 18 h and then lysed. Cells were lysed in lysis buffer [50 mM Hepes (pH 7.5), 1% (v/v) Igepal-C630, 1 mg/ml bacitracin, 1 mM EDTA, 10 mM NaF, 1 mM sodium orthovanadate, 10% (v/v) glycerol, 50 mM NaCl, 1 mM PMSF and Protease Inhibitor Cocktail Set III (Calbiochem)]. The detergent-insoluble materials were sedimented by centrifugation at 1600 g for 15 min at 4 °C. Cell lysate (500 μg) was used to precipitate FRS2 using an anti-FRS2 antibody. Immunoblots were visualized with ECL (enhanced chemiluminescence). Images were analysed with a Fuji LAS-1000 luminescent image analyser and quantified with Fuji Image Gauge software.

Cell imaging

Cells were seeded on to glass coverslips 24–48 h prior to stimulation. Serum-starved cells were stimulated with 10 ng/ml FGF9 for the indicated time periods. Stimulation was stopped by transfer of the coverslip to 4% (w/v) paraformaldehyde (pH 8.0). The pH in all subsequent steps was kept above 8.0. For antibody labelling, cells were permeabilized with 1% saponin in PBS for 5 min at 4 °C, washed with PBS and incubated with blocking solution [3% BSA/1% saponin/5% FCS in TBS (Tris-buffered saline; pH 8)] for 30 min at room temperature (20–22 °C). The Cy3-labelled antibody was diluted in incubation buffer (3% BSA/1% saponin in TBS) and incubated overnight at 4 °C. After washing in TBS 4–5 times, the coverslip was mounted on to a slide with mounting medium (0.1% p-phenylenediamine/50% glycerol in PBS at pH 7.5–8.0).

Microscopy

Confocal laser microscopy was performed with a Leica TCS SP system with a 63× oil-immersion objective. GFP was excited with an argon-visible light laser tuned to 488 nm, RFP and Cy3 were excited with a krypton laser tuned to 568 nm. GFP and Cy3 or RFP were detected via 514/10 nm and 595/10 nm band selection respectively. Fluorescence images were collected using a photomultiplier tube interfaced to an Intel Pentium II system running Leica TCS NT control software. Co-localization graphs were generated using the quantification function of the Leica NT software.

FLIM (fluorescence lifetime imaging microscopy)

Experiments were performed using an inverted confocal microscope (Leica TCS SP2) which was adapted for TCSPC (time-correlated single-photon counting) FLIM with a Becker and Hickl SPC 830 card using 64 or 256 time channels in a 3 GHz, Pentium IV, 1GB RAM computer running Windows XP. The samples were excited with a fs titanium sapphire laser (Coherent Mira, repetition rate 76 MHz.), pumped by a 6.5 W solid-state laser (Coherent Verdi V6). Images were obtained with a line scan speed of 200 Hz, with an image size of 512×512 pixels. We used a 63× water-immersion objective (numerical aperture NA=1.2) On the FLIM system the pixels were reduced to 256×256. The wavelength used for two-photon excitation was 900 nm, and the fluorescence was detected through a 525±25 nm interference filter using a cooled PMC100-01 detector (Becker and Hickl; based on a Hamamatsu H5772P-01 photomultiplier). The fluorescence decays were fitted with a single exponential decay model using Becker and Hickl's SPCImage software v2.8.3, and the GFP fluorescence lifetimes were displayed in a false colour map.

RESULTS

Expression and characterization of functional GFP-tagged FGFR2 receptors

To identify a suitable model system, a number of cell lines, including some osteoblastic and fibroblastic cells, were screened. Two criteria were stipulated for the cell line: (i) there should be no endogenous FGFR2 expression, and (ii) it must be able to express a sufficient quantity of transfected protein for high-quality microscopic imaging. Although a number of cell lines met the first criterium, they failed to stably express sufficient levels of exogenously introduced FGFR2. HEK-293T cells proved to be an ideal model cell line since they lack endogenous FGFR2 and they express a sufficient quantity of tagged FGFR2 allowing cellular imaging data to be obtained with clarity. Stable cell lines overexpressing wild-type, S252W or P253R FGFR2 fused to GFP (hereafter referred to as WTFGFR2, SWFGFR2 and PRFGFR2 respectively) were established. Fluorescence images of cells showed that all three fusion proteins were found predominantly in the plasma membrane, confirming that the GFP fusion did not interfere with receptor localization (results not shown). There is a possibility that the GFP fusion construct and/or overexpression of the FGFR2 in the HEK-293T cells induces dimerization of the receptor molecules and thereby constitutively activates the receptor. However, since we were comparing the differences in signalling between the wild-type and the two mutants, we expected that any dimerization or constitutive activation effect would apply equally to all three receptors. To ensure that any observed differences in intracellular signalling were due to the receptor mutation, cell lines with equivalent levels of receptor expression were obtained. A number of individual cell clones from each receptor cell pool were obtained by limiting dilution and analysed by immunoblotting. Three individual cell lines expressing the wild-type and the two mutant receptors were probed with an antibody against GFP and showed an equivalent level of receptor expression on SDS/PAGE gels (Figure 1A, upper panel). Furthermore, multiple clones of each receptor were analysed for the downstream MAPK response to rule out any clonal genotypic variation between the individual cell clones (results not shown). The receptors with extracellular point mutations showed similar migration patterns to that of the wild-type, except that the major bands were gel-shifted to a slightly higher molecular mass. Treatment of cells with SU5402 (a specific inhibitor of FGFR kinase [25]) inhibited the FGFR2 function, but had a negligible effect on the receptor mobility on SDS/PAGE, suggesting that the observed gel-shift with the mutant receptors was not due to increased phosphorylation (results not shown). The apparent increase in molecular mass of the bands for the SWFGFR2 and PRFGFR2 compared with wild-type may indicate an additional receptor glycosylation. Inhibition of glycosylation using tunicamycin results in all three receptors adopting a single identical molecular mass band on SDS/PAGE (Figure 1B). We evaluated and quantified the extent of overexpression of these receptors in HEK-293T cells in comparison with the endogenous level of the FGFR2 expression in a rat osteosarcoma cell line ROS17/2.8 (Figure 1C). Serially diluted total lysates of HEK-293T cells overexpressing WTFGFR2 were loaded with the ROS17/2.8 cell lysate and immunoblotted using an anti-FGFR2 antibody. The band intensities were quantified using densitometry which suggests that the level of receptor expression in transfected cells was approx. 4-fold higher than in ROS17/2.8 cells.

Expression of a functional GFP-tagged WTFGFR2, SWFGFR2 and PRFGFR2

Figure 1
Expression of a functional GFP-tagged WTFGFR2, SWFGFR2 and PRFGFR2

(A) Upper panel: an immunoblot of the wild-type (WT) and the Apert mutants [S252W (SW) and P253R (PR)] from two independently lysed cell extracts of the same clones analysed with an anti-GFP antibody (lanes 1, 2 and 3 and lanes 4, 5 and 6 respectively). Lower panel: the immunoblot was stripped and re-probed for Grb2 (a ubiquitous adapter protein in these cells) and shows equivalent levels of Grb2 expression. (B) Serum-starved cells were incubated with either 3 or 10 ng/ml tunicamycin overnight, cell lysates were prepared and immunoblotted (Blot) with an anti-GFP antibody. w.c.l., whole cell lysate (C) Comparative analysis of the level of GFP-tagged receptors overexpressed in HEK-293T cells by immunoblotting (IB). Total cell lysates (200 ng) from ROS17/2.8 cells (ROS) with the same amount of lysates serially diluted from WTFGFR2–GFP-expressing cells were loaded on to the gel and immunoblotted with an anti-FGFR2 antibody (upper panel) and then re-probed with an anti-tubulin antibody (lower panel). (D) Quantitative ligand-stimulated tyrosine phosphorylation of the wild-type (WT) and the Apert mutant receptors (SW and PR) indicating the kinase activity of each receptor from three independent experiments. The receptors were immunoprecipitated from 0.5 mg of total cell lysates using 10 μg of an anti-GFP antibody and analysed in an immunoblot with the anti-phosphotyrosine antibody. The phosphorylated receptor bands were quantified using densitometry where the unstimulated wild-type receptor phosphorylation was normalized as 1.0 and the relative increase or decrease in band intensities are shown. Values are means±S.D. (E) Serum-starved cells were stimulated with 10 ng/ml FGF9 and cell lysates were prepared. Total cell lysates of cells expressing the wild-type (WT) and the mutants were analysed in an immunoblot (IB) with the anti-phosphotyrosine antibody (upper panel), stripped and re-probed with an anti-FRS2 antibody (middle panel) and an anti-FGFR2 antibody (lower panel). (F) A comparison of the unstimulated and FGF9-stimulated receptor phosphorylation between the wild-type FGFR2 (WT) and wild-type with a C-terminal deletion similar to the K-sam-IIC3 (WTΔC) in stably transfected HEK-293T cells. Serum-starved cells expressing the wild-type FGFR2 and WTΔC were stimulated with 10 ng/ml FGF9 for 15 min. Cell lysates were prepared and immunoblotted (IB) using an anti-phosphotyrosine antibody (upper panel). The band intensity was normalized to the unstimulated WT lane and the relative changes are shown. The immunoblot was re-probed with an anti-GFP antibody (middle panel). A longer exposure of the anti-phosphotyrosine blot is shown in the lower panel to demonstrate that the middle receptor band is less phosphorylated. Molecular-mass standards are indicated on the left-hand side of (A), (B) and (E).

Figure 1
Expression of a functional GFP-tagged WTFGFR2, SWFGFR2 and PRFGFR2

(A) Upper panel: an immunoblot of the wild-type (WT) and the Apert mutants [S252W (SW) and P253R (PR)] from two independently lysed cell extracts of the same clones analysed with an anti-GFP antibody (lanes 1, 2 and 3 and lanes 4, 5 and 6 respectively). Lower panel: the immunoblot was stripped and re-probed for Grb2 (a ubiquitous adapter protein in these cells) and shows equivalent levels of Grb2 expression. (B) Serum-starved cells were incubated with either 3 or 10 ng/ml tunicamycin overnight, cell lysates were prepared and immunoblotted (Blot) with an anti-GFP antibody. w.c.l., whole cell lysate (C) Comparative analysis of the level of GFP-tagged receptors overexpressed in HEK-293T cells by immunoblotting (IB). Total cell lysates (200 ng) from ROS17/2.8 cells (ROS) with the same amount of lysates serially diluted from WTFGFR2–GFP-expressing cells were loaded on to the gel and immunoblotted with an anti-FGFR2 antibody (upper panel) and then re-probed with an anti-tubulin antibody (lower panel). (D) Quantitative ligand-stimulated tyrosine phosphorylation of the wild-type (WT) and the Apert mutant receptors (SW and PR) indicating the kinase activity of each receptor from three independent experiments. The receptors were immunoprecipitated from 0.5 mg of total cell lysates using 10 μg of an anti-GFP antibody and analysed in an immunoblot with the anti-phosphotyrosine antibody. The phosphorylated receptor bands were quantified using densitometry where the unstimulated wild-type receptor phosphorylation was normalized as 1.0 and the relative increase or decrease in band intensities are shown. Values are means±S.D. (E) Serum-starved cells were stimulated with 10 ng/ml FGF9 and cell lysates were prepared. Total cell lysates of cells expressing the wild-type (WT) and the mutants were analysed in an immunoblot (IB) with the anti-phosphotyrosine antibody (upper panel), stripped and re-probed with an anti-FRS2 antibody (middle panel) and an anti-FGFR2 antibody (lower panel). (F) A comparison of the unstimulated and FGF9-stimulated receptor phosphorylation between the wild-type FGFR2 (WT) and wild-type with a C-terminal deletion similar to the K-sam-IIC3 (WTΔC) in stably transfected HEK-293T cells. Serum-starved cells expressing the wild-type FGFR2 and WTΔC were stimulated with 10 ng/ml FGF9 for 15 min. Cell lysates were prepared and immunoblotted (IB) using an anti-phosphotyrosine antibody (upper panel). The band intensity was normalized to the unstimulated WT lane and the relative changes are shown. The immunoblot was re-probed with an anti-GFP antibody (middle panel). A longer exposure of the anti-phosphotyrosine blot is shown in the lower panel to demonstrate that the middle receptor band is less phosphorylated. Molecular-mass standards are indicated on the left-hand side of (A), (B) and (E).

To establish that the expressed receptors were functional as TKs, the wild-type and the mutant receptors were immunoprecipitated from FGF9-stimulated cells using an anti-GFP antibody, immunoblotted with an anti-phosphotyrosine antibody and quantified by densitometry (Figure 1D). Although all three cell lines were serum-starved under the same conditions and appear to express similar levels of receptors, cells expressing the WTFGFR2 and PRFGFR2 exhibited a higher level of basal receptor tyrosine phosphorylation than those with SWFGFR2. This suggests that the S252W point mutation somehow impedes background phosphorylation. The observed high level of basal receptor phosphorylation in WTFGFR2 and PRFGFR2 is not due to the GFP fusion, since transiently expressed non-tagged receptors portrayed a similar basal phosphorylation pattern (results not shown).

We next compared the effects of the expression of wild-type and the two mutant receptors on cellular tyrosine phosphoproteins. Immunoblotting with an anti-phosphotyrosine antibody revealed that in cells expressing the S252W and the P253R receptors, FRS2 underwent robust tyrosine phosphorylation (Figure 1E, top panel). The phosphorylation of FRS2 was minimal in cells expressing the wild-type FGFR2. In the S252W cells, FRS2 phosphorylation was apparent at 30 min post-stimulation, whereas in the P253R cells the phosphorylation can be seen within 15 min of stimulation. Re-probing of the blot with an anti-FRS antibody showed that, although no tyrosine phosphorylation was observed in the cell expressing WTFGFR2, there was a stimulation-dependent gel-shift of FRS2.

The ligand-dependent increase in protein phosphorylation is indicative of receptor kinase activity. Thus, due to higher-affinity/prolonged binding of the FGF ligand, it might be expected that more mutant receptor molecules at the cell surface were engaged and activated by the ligand, thereby causing an overall increase in the observed receptor phosphorylation. This is indeed true for the SWFGFR2 (Figure 1D); however, both the wild-type and the P253R receptors showed similar levels of ligand-stimulated receptor phosphorylation and hence kinase activity (Figure 1D). So despite also having an increased affinity for FGF ligand, the P253R mutant does not portray the same ligand-stimulated increase in receptor phosphorylation as the S252W receptor. This might be explained by the observation that the P253R mutant showed high levels of basal phosphorylation which could counteract the effect of increased affinity. The mutations therefore appeared to have an effect on both basal and ligand-stimulated receptor phosphorylation in these cells. We currently hypothesize that this may result from effects on the C-terminus of the wild-type receptor, since a deletion of the wild-type receptor (similar to the K-sam-IIC3 deleted receptor [26]) resulted in a drastic reduction of the basal receptor phosphorylation (Figure 1F). The C-terminus of the FGFR2 contains several non-conserved tyrosine residues that may act as auto-substrate. Interestingly, the C-terminal deletion resulted in a lower level of tyrosine phosphorylation in the middle receptor band which accounted for most of the tyrosine phosphorylation in the absence of ligand in the wild-type receptors (Figure 1F, exposure 2). The lack of the intermediate receptor band phosphorylation of the C-terminal wild-type deletion (WTΔC) is very similar to that which is seen for the S252W receptor.

Spatial/temporal distribution of the tyrosine-phosphorylated receptor and cellular phosphoproteins

To assess the effect of the extended engagement of FGFR2 by FGF on receptor kinase activity we investigated differences in spatial and temporal distribution of tyrosine-phosphorylated proteins. Serum-starved cells were stimulated over increasing periods of time with FGF9, fixed, permeabilized, stained with an anti-pY antibody coupled to Cy3, and observed by confocal microscopy. FGF9 was used in these experiments (as opposed to FGF2) because this ligand has been reported to be more specific for FGFR2 (IIIc) [15,27,28]. The specific nature of FGF9 for FGFR2 can also be exemplified in PC12 cells. These endogenously express FGFR1 and can be differentiated to form neurite extensions by FGF1 or FGF2 stimulation but not with FGF9. Exogenous expression of FGFR2 in these cells and the subsequent FGF9 stimulation results in neuronal differentiation. This suggests that this growth factor can be used as a specific ligand for FGFR2 activation [29]. Cellular images of the wild-type and mutant receptors, and total tyrosine-phosphorylated protein pre- and post-FGF9 stimulation are shown in Figure 2. Prior to stimulation, WTFGFR2 resided predominantly in the plasma membrane and limited amounts of tyrosine phosphorylated proteins co-localized with the receptor (Figure 2A, WTFGFR2 basal image and graph). Although this level of basal receptor phosphorylation appears to be contradictory to the observation in Figure 1(C), closer inspection of all of the sections of the cell image revealed a large amount of receptor co-localized with tyrosine-phosphorylated protein in sections closest to the glass slide. At 5 min after the addition of FGF9, some intermittent intracellular vesicular clusters containing receptors that were tyrosine phosphorylated, or localized with phosphorylated protein were observed (results not shown). This population increased at 15 min (Figure 2A, WTFGFR2 15 min image). A graphical representation of the fluorescence intensity through a cross-section of the cell confirmed that the receptor on the plasma membrane at 15 min post-stimulation was minimally tyrosine phosphorylated and/or did not co-localize with cellular tyrosine phosphorylated proteins. The presence of unphosphorylated receptor can be seen as a single green peak (i.e. without the equivalent red peak corresponding to the phosphorylated protein) at the point where the line drawn in the overlaid image crosses the membrane (Figure 2A, 15 min graphs). This demonstrates that phosphorylated receptor appears to be rapidly internalized post-stimulation. At 60 min the receptors that were found at the cell membrane appeared to be either tyrosine phosphorylated or co-localized with tyrosine phosphoproteins (Figure 2A, WTFGFR2 60 min image). As a control experiment, cells were stimulated with 100 ng/ml insulin. This showed little effect on the FGFR2 receptor localization, but widespread phosphoprotein-containing vesicles were observed (Figure 2B).

FGF9 induced co-localization of tyrosine-phosphorylated proteins with the wild-type and mutant FGFR2

Figure 2
FGF9 induced co-localization of tyrosine-phosphorylated proteins with the wild-type and mutant FGFR2

Stable HEK-293T cells expressing GFP-tagged wild-type and mutant FGFR2 were grown on a coverslip, serum-starved overnight and stimulated with 10 ng/ml FGF9 for the indicated time periods. Cells were fixed and counter-stained for tyrosine phosphorylated proteins with a pY99-Cy3 antibody. The cellular localization of tyrosine-phosphorylated proteins and GFP-tagged receptors were captured using a confocal microscope. (A) Localization of the WTFGFR2, SWFGFR2 and PRFGFR2 receptors with tyrosine-phosphorylated proteins. (B) A control experiment showing the co-localization of the WTFGFR2 with tyrosine-phosphorylated proteins following 15 min of 100 nM insulin stimulation. The confocal image demonstrates that the pY-labelled Cy3 antibody is sufficient to detect tyrosine-phosphorylated protein in the cell. Graphs are generated using Leica confocal software and are the representations of the fluorescent intensity of each pixel along the drawn line segment.

Figure 2
FGF9 induced co-localization of tyrosine-phosphorylated proteins with the wild-type and mutant FGFR2

Stable HEK-293T cells expressing GFP-tagged wild-type and mutant FGFR2 were grown on a coverslip, serum-starved overnight and stimulated with 10 ng/ml FGF9 for the indicated time periods. Cells were fixed and counter-stained for tyrosine phosphorylated proteins with a pY99-Cy3 antibody. The cellular localization of tyrosine-phosphorylated proteins and GFP-tagged receptors were captured using a confocal microscope. (A) Localization of the WTFGFR2, SWFGFR2 and PRFGFR2 receptors with tyrosine-phosphorylated proteins. (B) A control experiment showing the co-localization of the WTFGFR2 with tyrosine-phosphorylated proteins following 15 min of 100 nM insulin stimulation. The confocal image demonstrates that the pY-labelled Cy3 antibody is sufficient to detect tyrosine-phosphorylated protein in the cell. Graphs are generated using Leica confocal software and are the representations of the fluorescent intensity of each pixel along the drawn line segment.

In cells expressing SWFGFR2 the receptor resides predominantly at the cell membrane with a low level of phosphorylation or co-localization with tyrosine phosphorylated proteins before stimulation (Figure 2A, basal image and graph). Following FGF9 stimulation, an increase in the phosphorylation of the receptor, or concentration of proximal tyrosine phosphorylated proteins on the membrane was observed with time. At 15 min post-stimulation it is apparent that the SWFGFR2 does not undergo the same degree of endocytosis as was seen with the wild-type receptors (seen in approx. 90% of cells) (Figure 2A, SWFGFR2 15 min). Some punctate clusters did appear at 60 min but far less than those observed with the wild-type receptor (Figure 2A, SWFGFR2 60min). Thus the S252W mutation seems to restrict and/or delay the normal endocytic process and thereby also interfere with endocytosis of phosphorylated protein. The fact that the vast majority of the receptor and tyrosine-phosphorylated protein localize to the plasma membrane as opposed to the intracellular vesicles seen with the wild-type receptor suggest a serious alteration of spatial, and to some extent temporal, signalling by the S252W mutant receptor.

Cells expressing PRFGFR2 showed some clustering of phosphorylated receptors or tyrosine-phosphorylated proteins at the cell membrane in the basal state (Figure 2A, PRFGFR2 basal graph). After FGF9 stimulation, receptors on the plasma membrane were tyrosine-phosphorylated or associated with phosphoprotein (Figure 2A, PRFGFR2 15min). Although there was a progressive time-dependent increase of receptor phosphorylation on the cell membrane, a number of intracellular vesicles contain receptors were also seen at 60 min (Figure 2A, PRFGFR2 60 min). This is similar to the S252W mutant where a delayed and limited number of receptor and phosphoprotein-containing vesicles was observed. This suggests that both mutants share a similar internalization pattern and therefore would generate similar signalling complexes primarily on the cell membrane that could give rise to an altered mechanism of signalling.

MAPK activation

Having observed the altered spatial-temporal patterns of receptor and cellular phosphoprotein localization between the mutant and the wild-type receptors, we chose to investigate the effect of these two mutations on downstream cellular responses. One of the downstream effectors of FGFRs is the MAPK, ERK, which plays an essential role in mitogenesis [1,30]. Thus differences in the level of ERK phosphorylation would provide an insight into how the altered spatial-temporal receptor distributions of S252W and P253R mutations affect downstream signalling events. Cells were serum-starved overnight and stimulated with 10 ng/ml FGF2 or FGF9. Cell lysates were immunoblotted against phosphorylated ERK (Figure 3A, upper panel), stripped and re-probed against total ERK (Figure 3A, lower panel). Phospho-ERK immunoblots from three independent experiments were quantified using densitometry. FGF2 and FGF9 (Figures 3B and 3C respectively) stimulation resulted in a time-dependent increase in the ERK activation in all three cell lines which was sustained over the 60 min time course. Cells expressing FGFR2 with mutations displayed enhanced FGF2- and FGF9-stimulated ERK phosphorylation in comparison with the wild-type. Similar ERK phosphorylation patterns were observed in FGF2-stimulated ROS17/2.8 cells and FGF9-stimulated PC12 cells overexpressing the wild-type and mutant receptors (results not shown). This suggests that the observed altered receptor distribution in Figure 2 has a profound effect on downstream receptor signalling as demonstrated by the enhanced MAPK response. This observation underscores the pivotal role of receptor downstream responses, which has been demonstrated in a recent study with a mouse model of Apert's syndrome, where inhibition of MAPK relieved craniosynostosis and premature fusion of coronal suture and restored normal body mass [31].

Time course comparison of ERK activation between WTFGFR2, SWFGFR2 and PRFGFR2 cells

Figure 3
Time course comparison of ERK activation between WTFGFR2, SWFGFR2 and PRFGFR2 cells

Serum-starved cells expressing WTFGFR2, SWFGFR2 and PRFGFR2 (P253R) were stimulated with either 10 ng/ml FGF2 or FGF9 for the designated time period. Cells were washed once with PBS and lysed with lysis buffer. (A) Soluble protein (5 μg of whole cell lysates) was immunoblotted with an antibody that detects phosphorylated ERK (pERK; upper panel). The numbers represent the relative band intensity after normalization to wild-type (WT) at 60 min (denoted as 1.0). The blot was stripped and reprobed with an anti-ERK antibody (lower panel). (B) Densitometry quantification of the FGF2-stimulated ERK phosphorylation from three independent experiments after normalization to wild-type at 60 min as above. WT, WTFGFR2; SW, SWFGFR2; PR, PRFGFR2. Values are means±S.D. (C) Densitometry quantification of the FGF9-stimulated ERK phosphorylation from three independent experiments. WT, WTFGFR2; SW, SWFGFR2; PR, PRFGFR2. Values are means±S.D.

Figure 3
Time course comparison of ERK activation between WTFGFR2, SWFGFR2 and PRFGFR2 cells

Serum-starved cells expressing WTFGFR2, SWFGFR2 and PRFGFR2 (P253R) were stimulated with either 10 ng/ml FGF2 or FGF9 for the designated time period. Cells were washed once with PBS and lysed with lysis buffer. (A) Soluble protein (5 μg of whole cell lysates) was immunoblotted with an antibody that detects phosphorylated ERK (pERK; upper panel). The numbers represent the relative band intensity after normalization to wild-type (WT) at 60 min (denoted as 1.0). The blot was stripped and reprobed with an anti-ERK antibody (lower panel). (B) Densitometry quantification of the FGF2-stimulated ERK phosphorylation from three independent experiments after normalization to wild-type at 60 min as above. WT, WTFGFR2; SW, SWFGFR2; PR, PRFGFR2. Values are means±S.D. (C) Densitometry quantification of the FGF9-stimulated ERK phosphorylation from three independent experiments. WT, WTFGFR2; SW, SWFGFR2; PR, PRFGFR2. Values are means±S.D.

Cellular localization of FRS2

FRS2 has been shown to play a major role in the recruitment of the Grb2–SOS complex to the FGFR1 and in the subsequent downstream activation of the MAPK pathway [11,32]. In order to understand the cellular signalling process further, we investigated whether the differences in MAPK activation observed for FGFR2 and mutants thereof arose as a result of differential recruitment of FRS2. This was assessed by comparing the localization of C-terminally RFP-tagged FRS2 in cells expressing GFP-tagged wild-type or mutant receptors (Figure 4). Prior to FGF stimulation, FRS2 showed perfect co-localization with the wild-type receptor on the cell membrane (Figure 4A, WTFGFR2 basal image and graph). This is in contrast with expression of RFP alone which showed diffused distribution throughout the cell (Figure 4B). At 15 min post-FGF9 stimulation, intracellular vesicles containing both FGFR2 and FRS2 were observed (Figure 4A, WTFGFR2). By 60 min in >90% of cells, the majority of the vesicles had dissipated and both FGFR2 and FRS2 re-appeared at the plasma membrane (Figure 4A, WTFGFR2 60 min image and graph). Therefore the WTFGFR2 and FRS2 undergo transient ligand-dependent endocytosis. As expected, in a control using insulin stimulation of these cells, little effect on the cellular localization of the FGFR2 and FRS2 was observed (Figure 4B).

The localization of FRS2 with the wild-type and mutant FGFR2 following FGF9 stimulation

Figure 4
The localization of FRS2 with the wild-type and mutant FGFR2 following FGF9 stimulation

Stable HEK-293T cells expressing GFP-tagged wild-type and mutant FGFR2 were transiently transfected with RFP-tagged FRS2. At 24 h post-transfection, cells were seeded on to coverslips and allowed to grow for a further 48 h. Cells were serum-starved overnight and stimulated with 10 ng/ml FGF9 for the indicated time, fixed and analysed by confocal microscopy. (A) Co-localization of the WTFGFR2–GFP, SWFGFR2–GFP and PRFGFR2–GFP with FRS2–RFP. (B) Co-localization of FRS2–RFP with WTFGFR2 and SWFGFR2 following 15 min insulin stimulation. Insulin has no effect on FRS2 localization and thus it is a specific substrate for FGFR2. Furthermore, a control experiment showed distinct pattern of expression of isolated RFP when compared with the pattern observed when RFP is tagged to FRS2.

Figure 4
The localization of FRS2 with the wild-type and mutant FGFR2 following FGF9 stimulation

Stable HEK-293T cells expressing GFP-tagged wild-type and mutant FGFR2 were transiently transfected with RFP-tagged FRS2. At 24 h post-transfection, cells were seeded on to coverslips and allowed to grow for a further 48 h. Cells were serum-starved overnight and stimulated with 10 ng/ml FGF9 for the indicated time, fixed and analysed by confocal microscopy. (A) Co-localization of the WTFGFR2–GFP, SWFGFR2–GFP and PRFGFR2–GFP with FRS2–RFP. (B) Co-localization of FRS2–RFP with WTFGFR2 and SWFGFR2 following 15 min insulin stimulation. Insulin has no effect on FRS2 localization and thus it is a specific substrate for FGFR2. Furthermore, a control experiment showed distinct pattern of expression of isolated RFP when compared with the pattern observed when RFP is tagged to FRS2.

In resting cells expressing the S252W mutant, both the receptor and FRS2 showed membrane co-localization (Figure 4A, SWFGFR2 basal image and graph). However, unlike in the wild-type cells, most FRS2 and SWFGFR2 molecules continued to reside on the cell membrane throughout the stimulation time period (Figure 4A, 15 and 60 min). This is consistent with the earlier observation of phosphorylated SWFGFR2 and/or another phosphorylated protein remaining on the membrane for an extend period of time (Figure 2A). Interestingly, FGF9 stimulation of SWFGFR2 induced the formation of intracellular vesicles containing only FRS2. This is in contrast with the wild-type receptor, suggesting distinct modes of FRS2 processing (Figure 4A, SWFGFR2 at 15 and 60 min graph, arrows showing intense FRS2–RFP peak only). The observed receptor-independent FRS2 vesicles may be caused by the inability of the S252W receptor to undergo prompt endocytosis.

The PRFGFR2 exhibits yet another apparent co-localization pattern with FRS2. In a similar manner to the WTFGFR2 and SWFGFR2, at the basal state, both the receptor and FRS2 co-localize at the plasma membrane (Figure 4A, PRFGFR2 basal image and graph). However, unlike the other receptors, very few intracellular vesicles containing FRS2 and the receptor (as seen with WTFGFR2), or FRS2 alone (as seen with the SWFGFR2) were formed throughout the period of stimulation (Figure 4A). The majority of the receptors and FRS2 remained on the membrane throughout the time course. This is consistent with the prolonged appearance of phosphorylated receptor (or recruited phosphoprotein) at the membrane after stimulation (Figure 2A, PRFGFR2). The changes that occur in the ligand–receptor interaction as a result of the mutations completely alter the cellular localization of FRS2. Since FRS2 is the main FGFR-docking protein responsible for MAPK activation, the changes in its cellular localization may contribute to the observed enhanced MAPK response shown in Figure 3.

Protein precipitation with FRS2

As the wild-type and mutants show distinct patterns of cellular localization with FRS2, we next investigated whether this had an impact on the assembly of the signalling complexes generated. FRS2 was immunoprecipitated from cells expressing the wild-type and the mutant FGFR2 using an anti-FRS2 antibody. The precipitants were subjected to immunoblotting with an anti-phosphotyrosine antibody (Figure 5A, upper panel), stripped and re-probed with an anti-FRS2 antibody (Figure 5A, upper-middle panel), anti-FGFR2 (Figure 5A, lower-middle panel) and IgG (Figure 5A, bottom panel). The unstimulated cells expressing the wild-type FGFR2 receptor showed very low levels of FRS2 phosphorylation and co-precipitation of receptors. This suggests that, although the proteins co-localize at the cell membrane (seen in the overlay of the cell images, Figure 4A), they may not be interacting or are in a complexed state. Following FGF stimulation, there is an increase in the level of FRS2 phosphorylation and co-precipitation of receptors (Figure 5A, lane 2). In addition, a number of cellular tyrosine-phosphorylated proteins are seen to be in complex with the receptor and FRS2, most notably a protein with an apparent molecular mass of 60 kDa. Thus in cells expressing the wild-type receptor it appears that the association of FRS2 and the FGFR2 predominantly occurs in a stimulation-dependent manner.

Co-precipitation of FGFR2 with FRS2

Figure 5
Co-precipitation of FGFR2 with FRS2

(A) Serum-starved HEK-293T cells expressing WTFGFR2 (WT), SWFGFR2 (SW) and PRFGFR2 (PR) were stimulated with 10 ng/ml FGF9 and whole cell lysates were prepared. FRS2 was precipitated using an anti-FRS2 antibody (IP: Anti-FRS2) and immunoblotted with an anti-phosphotyrosine antibody (IB: Anti-pY) (upper panel), stripped and re-probed with an anti-FRS2 antibody (IB: Anti-FRS2; upper-middle panel), anti-FGFR2 (IB: Anti-FGFR2; lower-middle panel) and IgG levels are shown in the bottom panel. ‘−’ denotes unstimulated and ‘+’ denotes 15 min FGF9 stimulation. The numbers denote the band intensity normalized to 1.0 using the wild-type stimulated lane. (B) FRS2 was immunoprecipitated with GST-p13suc1 (IP: p13suc-1) from unstimulated (B), FGF2- and FGF9-stimulated ROS17/2.8 cells and probed with an anti-phosphotyrosine antibody (IB: Anti-pY; upper panel). The blot was stripped and re-probed with an anti-FGFR2 antibody (IB: Anti-FGFR2; upper-middle panel), an anti-FRS2 antibody (IB: Anti-FRS2; lower middle panel) and an anti-GST antibody (IB: Anti-GST; lower panel).

Figure 5
Co-precipitation of FGFR2 with FRS2

(A) Serum-starved HEK-293T cells expressing WTFGFR2 (WT), SWFGFR2 (SW) and PRFGFR2 (PR) were stimulated with 10 ng/ml FGF9 and whole cell lysates were prepared. FRS2 was precipitated using an anti-FRS2 antibody (IP: Anti-FRS2) and immunoblotted with an anti-phosphotyrosine antibody (IB: Anti-pY) (upper panel), stripped and re-probed with an anti-FRS2 antibody (IB: Anti-FRS2; upper-middle panel), anti-FGFR2 (IB: Anti-FGFR2; lower-middle panel) and IgG levels are shown in the bottom panel. ‘−’ denotes unstimulated and ‘+’ denotes 15 min FGF9 stimulation. The numbers denote the band intensity normalized to 1.0 using the wild-type stimulated lane. (B) FRS2 was immunoprecipitated with GST-p13suc1 (IP: p13suc-1) from unstimulated (B), FGF2- and FGF9-stimulated ROS17/2.8 cells and probed with an anti-phosphotyrosine antibody (IB: Anti-pY; upper panel). The blot was stripped and re-probed with an anti-FGFR2 antibody (IB: Anti-FGFR2; upper-middle panel), an anti-FRS2 antibody (IB: Anti-FRS2; lower middle panel) and an anti-GST antibody (IB: Anti-GST; lower panel).

In unstimulated SWFGFR2-expressing cells, FRS2 is in complex with a number of highly phosphorylated cellular proteins (Figure 5A, upper panel, lane 3). However, following FGF9 stimulation, the amounts of cellular phosphoproteins in complex with FRS2 and co-precipitation of the receptor shows a slight reduction (Figure 5A, upper panel, lane 4). This is different from the observation with the wild-type receptor, where ligand stimulation led to an increase in tyrosine-phosphorylated protein recruited to the receptor–FRS2 complex. However, the phosphorylation of the apparent 60 kDa protein increases with FGF stimulation. In cells expressing the P253R mutant receptors, phosphorylation of FRS2 remains virtually unchanged between unstimulated cells and FGF-stimulated cells. Although a slight reduction in FGF-stimulated complex formation by cellular phosphoproteins is seen, an almost equal level of receptors were co-precipitated from both unstimulated and FGF9-stimulated cells (Figure 5A, upper panel, lanes 5 and 6). Therefore the P253R receptors appear to be constitutively associated or in a multi-protein complex with FRS2. The stimulation-dependent FRS2 tyrosine phosphorylation appears to be more robust in the mutants (Figure 5A, upper panel, and Figure 1E).

To confirm the observation of a lack of constitutive association of FGFR2 and FRS2 we also tested the stimulation-dependent association of FGFR2 with FRS2 in ROS17/2.8 cells. These cells express endogenous FGFR2. Serum-starved cells were stimulated with FGF2 and FGF9 and cell lysates were prepared and pulled down with GST-p13suc1. The pulldown experiment was performed because the FRS2 antibody performed poorly in ROS cells. The precipitants were subjected to immunoblotting with an anti-phosphotyrosine antibody (Figure 5B, upper panel), stripped and re-probed with an anti-FGFR2 (Figure 5B, upper-middle panel), anti-FRS2 (Figure 5B, lower-middle panel) and anti-GST antibodies (Figure 5B, bottom panel). There was an apparent stimulation-dependent receptor co-precipitation with FRS2 similar to the observation in FGFR2-transfected HEK-293T cells.

Direct interaction of FRS2 with the wild-type and mutant receptors

Although immunoprecipitation can be an indication of direct interaction, co-precipitation can also occur due to the formation of ternary complexes with other cellular proteins. Since the FRS2 pulldown and the co-localization studies suggested differences in FRS2 recruitment to the receptor, FLIM was chosen to unambiguously investigate whether these differences were the result of altered direct interaction between these two proteins [24,33,34]. FLIM has successfully been used to show direct interactions of proteins in cells [24]. Upon excitation, if the donor GFP (which is fused to FGFR2) is within the critical distance of 10 nm of an acceptor RFP (which is tagged to a target protein, i.e. FRS2), FRET (Förster resonance energy transfer) occurs from the donor to the acceptor. This results in a reduction of the emission lifetime of the donor molecule (GFP) which is measurable and indicates a direct interaction between the two molecules. The use of FLIM to detect FRET can be very accurate and has the advantage that only the GFP donor fluorescence needs to be measured. At basal state there are a few isolated clusters in which interaction between the WTFGFR2 and FRS2 can be observed (Figure 6A, basal lifetime image). This low level of FRS2–FGFR2 interaction seems to have little effect on the average lifetime graph where the peak is centred on approx. 2.0 ns (similar to that for isolated GFP; Figures 6A, basal and 6D) and cells expressing only the GFP-tagged wild-type receptor (results not shown). On FGF9 stimulation, a large population of punctate intracellular vesicle clusters show very short lifetimes, indicative of a direct interaction (Figure 6A, 15 min lifetime graph; peak shifts to left). This corroborates previously described evidence for rapid dissipation of receptor and FRS2 from the membrane on stimulation (Figures 2A and 4A respectively). FRS2 and FGFR2 appear to be no longer engaged in direct interaction at 60 min (Figure 6A). Therefore from these results we can conclude that the wild-type receptor interacts with FRS2 transiently in vesicles in a receptor activation-dependent manner and the interaction eventually dissociates with sustained ligand activation.

Comparison of the direct interaction between FRS2 and the receptors using FLIM

Figure 6
Comparison of the direct interaction between FRS2 and the receptors using FLIM

Stable HEK-293T cells expressing GFP-tagged wild-type and mutant FGFR2 were transiently transfected with RFP-tagged FRS2. At 24 h post-transfection, cells were seeded on to coverslips and allowed to grow for a further 48 h. Cells were serum-starved overnight and stimulated with 10 ng/ml FGF9 for the indicated time periods, fixed and analysed by confocal microscopy. Co-localization of FRS2–RFP with the wild-type and the mutants were acquired. The lifetime measurements were made over a 300 s acquisition period as described in the Experimental section. The top panels show the fluorescence intensity image and the middle panels show the lifetime image mapped to an arbitrary colour scale from blue to red. The fluorescence lifetime distribution for all pixels in the field of view is provided on in the bottom panels with the same colour scale. (A) Co-localization and FLIM image of FRS2–RFP with WTFGFR2–GFP (WT). (B) Co-localization and FLIM image of FRS2–RFP with SWFGFR2-GFP (SW). (C) Co-localization and FLIM image of FRS2–RFP with PRFGFR2–GFP (PR). (D) The control lifetime of the GFP and RFP. Graphs represent the average lifetime of the entire population of cells seen in the image.

Figure 6
Comparison of the direct interaction between FRS2 and the receptors using FLIM

Stable HEK-293T cells expressing GFP-tagged wild-type and mutant FGFR2 were transiently transfected with RFP-tagged FRS2. At 24 h post-transfection, cells were seeded on to coverslips and allowed to grow for a further 48 h. Cells were serum-starved overnight and stimulated with 10 ng/ml FGF9 for the indicated time periods, fixed and analysed by confocal microscopy. Co-localization of FRS2–RFP with the wild-type and the mutants were acquired. The lifetime measurements were made over a 300 s acquisition period as described in the Experimental section. The top panels show the fluorescence intensity image and the middle panels show the lifetime image mapped to an arbitrary colour scale from blue to red. The fluorescence lifetime distribution for all pixels in the field of view is provided on in the bottom panels with the same colour scale. (A) Co-localization and FLIM image of FRS2–RFP with WTFGFR2–GFP (WT). (B) Co-localization and FLIM image of FRS2–RFP with SWFGFR2-GFP (SW). (C) Co-localization and FLIM image of FRS2–RFP with PRFGFR2–GFP (PR). (D) The control lifetime of the GFP and RFP. Graphs represent the average lifetime of the entire population of cells seen in the image.

A high level of basal interaction was observed between SWFGFR2 and FRS2 (blue colouration in the lifetime image and the shorter average lifetime of the population at 1.7 ns; Figure 6B). This level of association appears to be partly lost on FGF9 stimulation (average lifetime closer to 2.0 ns, i.e. graph shifts to the right; Figure 6B, 15 and 60 min). This observation correlates with the anti-FRS2 antibody immunoprecipitation experiment (Figure 5A) where ligand stimulation showed a decrease in the SWFGFR2 and cellular phosphoprotein co-precipitation with FRS2. The loss of direct FRS2 association with the S252W receptor upon FGF9 stimulation further correlates with the previously observed stimulation-dependent sequestering of FRS2 into independent vesicles (Figure 4A). Therefore, in a similar manner to FGFR1, FRS2 is constitutively associated with SWFGFR2; however, ligand stimulation of the mutant receptor leads to the dissociation of this interaction. This is the opposite of the observation in cells expressing the wild-type receptors where ligand stimulation leads to the formation of FRS2–receptor interactions.

The P253R mutant, like the S252W receptor, showed direct interactions between the receptor and FRS2 at the basal level. However, unlike the S252W receptor but similar to the wild-type receptor, FGF9 stimulation results in an increase in P253R–FRS2 interaction (graph peak shifts to the left; Figure 6C). However, at 15 min all interactions between the P253R receptor and FRS2 occurred exclusively at the plasma membrane (Figure 6C, 15 min lifetime image and graph). This is again consistent with the previously observed co-localization of PRFGFR2 with FRS2 (Figure 4A). The combined FLIM data suggest that all three receptors are involved in profoundly different interactions with FRS2. The WTFGFR2 and PRFGFR2 display a stimulation-dependent interaction with FRS2 in the plasma and vesicular membranes respectively. The SWFGFR2, on the other hand, shows a direct constitutive interaction with FRS2 on the plasma membrane.

Wild-type FGFR2 localizes to the lysosome

Throughout the course of the present study it was apparent that the wild-type FGFR2 underwent ligand-stimulated receptor endocytosis into intracellular vesicles, whereas the mutant receptors were resistant to this process. This suggests that endocytosis into intracellular vesicles is a fundamental signalling difference between the wild-type and the two mutant receptors. Endocytosis plays a pivotal role in activation or down-regulation of many receptor TKs. In an effort to identify the intracellular vesicle to which FGFR2 is processed, we probed cells with various vesicle markers including those for endosomes, caveolae and lysosomes. The vast majority of the ligand-stimulated endocytosed receptors co-localized with a lysosomal marker (Figure 7) (it should be noted that the lysosomal marker used in the present study is a dye that detects low pH levels and therefore could also detect late endosomes). In unstimulated cells, the FGFR2 does not co-localize with the marker (Figures 7a, 7d and 7g); however, on FGF9 stimulation a number of vesicles containing the receptor were co-stained (Figures 7b, 7e and 7h). Further ligand stimulation resulted in virtually all of the receptors being identified as being localized with the lysosomal marker (Figures 7e, 7f and 7i). Lysosomes are specialized organelles containing degradative enzymes with an internal pH of 4.8. Therefore localization of the receptor to the lysosome would have one of two possible outcomes: degradation of the receptors by the resident proteases and/or dissociation of the ligand from the receptor owing to the low pH environment. Nevertheless, either process would result in receptor down-regulation. In contrast with the WTFGFR2, the two mutant receptors showed drastic reduction in their ability to endocytose to lysosomes, thus we would expect the observed relative enhanced MAPK response (Figure 3).

FGFR2 localizes to the lysosome upon FGF9 stimulation

Figure 7
FGFR2 localizes to the lysosome upon FGF9 stimulation

HEK-293T cells stably expressing wild-type FGFR2 were grown on a coverslip. Overnight serum-starved cells were incubated with LysoTracker Red DND99 dye for 15 min and then stimulated with 10 ng/ml FGF9 for the indicated time period and fixed. The cellular localization of GFP-tagged receptors and lysosomes are captured using confocal microscopy. The image represents the mid-section of the cells.

Figure 7
FGFR2 localizes to the lysosome upon FGF9 stimulation

HEK-293T cells stably expressing wild-type FGFR2 were grown on a coverslip. Overnight serum-starved cells were incubated with LysoTracker Red DND99 dye for 15 min and then stimulated with 10 ng/ml FGF9 for the indicated time period and fixed. The cellular localization of GFP-tagged receptors and lysosomes are captured using confocal microscopy. The image represents the mid-section of the cells.

DISCUSSION

Previous studies have suggested that prolonged receptor engagement and altered specificity for growth factors are the only outwardly discernable consequences of the S252W and P253R point mutations [15,16,35,36]. It is thus remarkable to observe such dramatic changes in the properties of the receptor molecules defined by clear differences in glycosylation, phosphorylation, direct interaction with FRS2 and downstream MAPK activity. In addition, bearing in mind the proximity of the point mutations and the structurally subtle nature of the changes which these produce in growth factor recognition, it is also intriguing to observe the profound changes in functional properties between SWFGFR2 and PRFGFR2 themselves.

The receptors appear as more than one band on an immunoblot, which most probably reflects differences in glycosylation. It is not clear mechanistically how this altered glycosylation might impinge on intracellular signalling, and there is no structural basis for glycosylation affecting the interface between FGF and FGFR2. Total ablation of glycosylation can result in an increase in ligand-independent dimerization and phosphorylation [37]. However, the wealth of data on the direct interaction between FGF and the extracellular domain of FGFR2 infer that the effect of N-glycosylation can largely be ignored [16].

One key feature of the point mutations in the receptors, which is independent of the time of growth factor engagement, is the observed difference in levels of basal phosphorylation. Although wild-type and PRFGFR2 are highly phosphorylated, SWFGFR2 is significantly less phosphorylated (Figure 1E). As a result, the effect of stimulation appears as a dramatic increase in phosphorylation of SWFGFR2; this is less profound on the other two receptors (Figure 1D). This demonstrates a clear phenotypic alteration between the mutant receptors themselves. A striking difference between the S252W and the P253R is that the S252W has an apparent higher molecular mass which could be viewed as a more ‘matured’ form and correlates with reduced basal phosphorylation. The wild-type and the P253R contain more of the intermediate form of the receptors which shows a higher level of basal phosphorylation. The C-terminus of the receptor appears to play a vital part in the ‘maturity’ of receptors. The C-terminal truncated receptor shows less basal receptor phosphorylation compared with the wild-type, suggesting that it might play a part in receptor processing (Figure 1F).

On stimulation, a further clear phenotypic difference between the cells expressing the different receptors was apparent. The absence of phosphorylated WTFGFR2 on the cell membrane (Figure 2) and its appearance in vesicles suggests that normal FGFR2 signalling involves endocytosis of the receptor once ligand is bound. In contrast, the phosphorylated mutant receptors and/or subsidiary phosphoproteins seem to remain at the membrane throughout the 60 min time course (Figure 2A). Thus the prolonged engagement of the receptors expressing the extracellular mutations confers some inhibitory effect on the endocytic process. Since ligand-stimulated internalization has been reported as playing a pivotal role in signal down-regulation for a number of other TK receptors, it is likely that the inability of the mutant receptors to undergo endocytosis may result in the apparent elevated levels of MAPK activity [4,3842].

The prolonged receptor engagement by ligand results in a dramatic change in complex formation between FRS2 and the receptor. In unstimulated cells, all three receptors appear to co-localize with FRS2 (Figure 4). However, immunoprecipitation assays and FLIM (Figures 5A and 6A respectively) reveal that direct interaction with WTFGFR2 occurs only in isolated clusters, suggesting that although they all have the same spatial distribution, the majority of the FRS2 is not interacting directly with the wild-type receptor. This is intriguing in light of the observation of a basal constitutive complex between FRS2 and FGFR1 [10], and suggests significant differences in early signalling complex protein recruitment between these receptors. Interestingly, very clear differences in direct interaction between the unstimulated mutant receptors and FRS2 were demonstrated by the lifetimes observed in the FLIM data, with SWFGFR2 being the most pronounced.

In cells expressing WTFGFR2, the phosphorylated FGFR2–FRS2 interaction occurs predominantly in the lysosome (Figures 6A and 7). The P253R mutant receptor complex prevails almost exclusively at the cell membrane (Figures 4 and 6). This is consistent with the observation of the distinct membrane localization of phosphorylated PRFGFR2 receptor and/or phosphoprotein on stimulation (Figure 2 and Figure 5A). The level of direct interaction of SWFGFR2 with FRS2 declines following stimulation compared with the other two receptors (Figure 6B); this is consistent with the immunoprecipitation experiment where the S252W receptor and phosphoprotein precipitated by FRS2 shows a stimulation-dependent decrease (Figure 5A). Since each of the three receptors show different properties of spatial and temporal distribution with FRS2, it is likely that time of engagement and the ensuing localization of the receptors dictates how signalling complexes are formed. The Apert syndrome mutant receptors in the HEK-293T cells provide a useful model system to show the dramatic effect of change in the temporal activity of signalling molecules. Indeed the present study represents the first demonstration of the effects of modulating the time of receptor engagement through an identical ligand. The resulting changes observed with respect to receptor phosphorylation and localization, protein recruitment and downstream MAPK response strongly suggest that the time of receptor engagement is a fundamental determinant of specificity in intracellular signal transduction [22]. It is clear from the present study that without invoking a time element to the definition of a signalling process, no real understanding of the ultimate outcome can be attained. Since the single point mutations result in a dramatic change to the basic protein signalling mechanism in distinct ways, the chances of predicting these from a global knowledge and quantification of the process for the wild-type are small. This argues that an holistic appreciation of TK-mediated signal transduction is likely to be a distant goal.

This work was funded by the Wellcome Trust. J. E. L is a Wellcome Trust Senior Research Fellow. A. C. S. is a Countess of Lisburne Scholar. We are indebted to J. Brockes and K. Bowers for their critical reading of the manuscript.

Abbreviations

     
  • DMEM

    Dulbecco's modified Eagle's medium

  •  
  • ERK

    extracellular-signal-regulated kinase

  •  
  • FBS

    foetal bovine serum

  •  
  • FGF

    fibroblast growth factor

  •  
  • FGFR

    FGF receptor

  •  
  • FLIM

    fluorescent lifetime imaging microscopy

  •  
  • FRET

    Förster resonance energy transfer

  •  
  • FRS2

    FGFR substrate 2

  •  
  • GFP

    green fluorescent protein

  •  
  • Grb2

    growth-factor-receptor-bound protein 2

  •  
  • GST

    glutathione transferase

  •  
  • HEK-293T cells

    human embryonic kidney-293 cells expressing the large T-antigen of SV40 (simian virus 40)

  •  
  • HSPG

    heparan sulfate proteoglycan

  •  
  • Ig

    immunoglobulin

  •  
  • MAPK

    mitogen-activated protein kinase

  •  
  • RFP

    red fluorescent protein

  •  
  • SOS

    Son of Sevenless

  •  
  • TBS

    Tris-buffered saline

  •  
  • TK

    tyrosine kinase

References

References
1
Powers
C. J.
McLeskey
S. W.
Wellstein
A.
Fibroblast growth factors, their receptors and signaling
Endocr. Relat. Cancer
2000
, vol. 
7
 (pg. 
165
-
197
)
2
ADHR Consortium
Autosomal dominant hypophosphataemic rickets is associated with mutations in FGF23
Nat. Genet.
2000
, vol. 
26
 (pg. 
345
-
348
)
3
Ornitz
D. M.
Itoh
N.
Fibroblast growth factors
Genome Biol.
2001
, vol. 
2
  
Reviews 3005
4
Yayon
A.
Klagsbrun
M.
Esko
J. D.
Leder
P.
Ornitz
D. M.
Cell surface, heparin-like molecules are required for binding of basic fibroblast growth factor to its high affinity receptor
Cell
1991
, vol. 
64
 (pg. 
841
-
848
)
5
Spivak-Kroizman
T.
Lemmon
M. A.
Dikic
I.
Ladbury
J. E.
Pinchasi
D.
Huang
J.
Jaye
M.
Crumley
G.
Schlessinger
J.
Lax
I.
Heparin-induced oligomerization of FGF molecules is responsible for FGF receptor dimerization, activation, and cell proliferation
Cell
1994
, vol. 
79
 (pg. 
1015
-
1024
)
6
Pellegrini
L.
Burke
D. F.
von Delft
F.
Mulloy
B.
Blundell
T. L.
Crystal structure of fibroblast growth factor receptor ectodomain bound to ligand and heparin
Nature
2000
, vol. 
407
 (pg. 
1029
-
1034
)
7
Schlessinger
J.
Plotnikov
A. N.
Ibrahimi
O. A.
Eliseenkova
A. V.
Yeh
B. K.
Yayon
A.
Linhardt
R. J.
Mohammadi
M.
Crystal structure of a ternary FGF–FGFR–heparin complex reveals a dual role for heparin in FGFR binding and dimerization
Mol. Cell
2000
, vol. 
6
 (pg. 
743
-
750
)
8
Mohammadi
M.
Dikic
I.
Sorokin
A.
Burgess
W. H.
Jaye
M.
Schlessinger
J.
Identification of six novel autophosphorylation sites on fibroblast growth factor receptor 1 and elucidation of their importance in receptor activation and signal transduction
Mol. Cell. Biol.
1996
, vol. 
16
 (pg. 
977
-
989
)
9
Schlessinger
J.
Cell signaling by receptor tyrosine kinases
Cell
2000
, vol. 
103
 (pg. 
211
-
225
)
10
Ong
S. H.
Guy
G. R.
Hadari
Y. R.
Laks
S.
Gotoh
N.
Schlessinger
J.
Lax
I.
FRS2 proteins recruit intracellular signaling pathways by binding to diverse targets on fibroblast growth factor and nerve growth factor receptors
Mol. Cell. Biol.
2000
, vol. 
20
 (pg. 
979
-
989
)
11
Kouhara
H.
Hadari
Y. R.
Spivak-Kroizman
T.
Schilling
J.
Bar-Sagi
D.
Lax
I.
Schlessinger
J.
A lipid-anchored Grb2-binding protein that links FGF-receptor activation to the Ras/MAPK signaling pathway
Cell
1997
, vol. 
89
 (pg. 
693
-
702
)
12
Ong
S. H.
Lim
Y. P.
Low
B. C.
Guy
G. R.
SHP2 associates directly with tyrosine phosphorylated p90 (SNT) protein in FGF-stimulated cells
Biochem. Biophys. Res. Commun.
1997
, vol. 
238
 (pg. 
261
-
266
)
13
Hadari
Y. R.
Kouhara
H.
Lax
I.
Schlessinger
J.
Binding of Shp2 tyrosine phosphatase to FRS2 is essential for fibroblast growth factor-induced PC12 cell differentiation
Mol. Cell. Biol.
1998
, vol. 
18
 (pg. 
3966
-
3973
)
14
Wong
A.
Lamothe
B.
Lee
A.
Schlessinger
J.
Lax
I.
FRS2α attenuates FGF receptor signaling by Grb2-mediated recruitment of the ubiquitin ligase Cbl
Proc. Natl. Acad. Sci. U.S.A.
2002
, vol. 
99
 (pg. 
6684
-
6689
)
15
Anderson
J.
Burns
H. D.
Enriquez-Harris
P.
Wilkie
A. O. M.
Heath
J. K.
Apert syndrome mutations in fibroblast growth factor receptor 2 exhibit increased affinity for FGF ligand
Hum. Mol. Genet.
1998
, vol. 
7
 (pg. 
1475
-
1483
)
16
Ibrahimi
O. A.
Eliseenkova
A. V.
Plotnikov
A. N.
Yu
K.
Ornitz
D. M.
Mohammadi
M.
Structural basis for fibroblast growth factor receptor 2 activation in Apert syndrome
Proc. Natl. Acad. Sci. U.S.A.
2001
, vol. 
98
 (pg. 
7182
-
7187
)
17
Yu
K.
Herr
A. B.
Waksman
G.
Ornitz
D. M.
Loss of fibroblast growth factor receptor 2 ligand-binding specificity in Apert syndrome
Proc. Natl. Acad. Sci. U.S.A.
2000
, vol. 
97
 (pg. 
14536
-
14541
)
18
Cohen
M. M.
Jr
Kreiborg
S.
Lammer
E. J.
Cordero
J. F.
Mastroiacovo
P.
Erickson
J. D.
Roeper
P.
Martinez-Frias
M. L.
Birth prevalence study of the Apert syndrome
Am. J. Med. Genet.
1992
, vol. 
42
 (pg. 
655
-
659
)
19
Oldridge
M.
Lunt
P. W.
Zackai
E. H.
Donald-McGinn
D. M.
Muenke
M.
Moloney
D. M.
Twigg
S. R.
Heath
J. K.
Howard
T. D.
Hoganson
G.
, et al. 
Genotype-phenotype correlation for nucleotide substitutions in the IgII-IgIII linker of FGFR2
Hum. Mol. Genet.
1997
, vol. 
6
 (pg. 
137
-
143
)
20
Oldridge
M.
Zackai
E. H.
Donald-McGinn
D. M.
Iseki
S.
Morriss-Kay
G. M.
Twigg
S. R.
Johnson
D.
Wall
S. A.
Jiang
W.
Theda
C.
, et al. 
De novo alu-element insertions in FGFR2 identify a distinct pathological basis for Apert syndrome
Am. J. Hum. Genet.
1999
, vol. 
64
 (pg. 
446
-
461
)
21
Wilkie
A. O. M.
Slaney
S. F.
Oldridge
M.
Poole
M. D.
Ashworth
G. J.
Hockley
A. D.
Hayward
R. D.
David
D. J.
Pulleyn
L. J.
Rutland
P.
, et al. 
Apert syndrome results from localized mutations of FGFR2 and is allelic with Crouzon syndrome
Nat. Genet.
1995
, vol. 
9
 (pg. 
165
-
172
)
22
O'Rourke
L.
Ladbury
J. E.
Specificity is complex and time consuming: mutual exclusivity in tyrosine kinase-mediated signaling
Acc. Chem. Res.
2003
, vol. 
36
 (pg. 
410
-
416
)
23
Ahmed
Z.
Beeton
C. A.
Williams
M. A.
Clements
D.
Baldari
C. T.
Ladbury
J. E.
Distinct spatial and temporal distribution of ZAP70 and Lck following stimulation of interferon and T-cell receptors
J. Mol. Biol.
2005
, vol. 
353
 (pg. 
1001
-
1010
)
24
Peter
M.
Ameer-Beg
S. M.
Hughes
M. K.
Keppler
M. D.
Prag
S.
Marsh
M.
Vojnovic
B.
Ng
T.
Multiphoton-FLIM quantification of the EGFP-mRFP1 FRET pair for localization of membrane receptor-kinase interactions
Biophys. J.
2005
, vol. 
88
 (pg. 
1224
-
1237
)
25
Mohammadi
M.
McMahon
G.
Sun
L.
Tang
C.
Hirth
P.
Yeh
B. K.
Hubbard
S. R.
Schlessinger
J.
Structures of the tyrosine kinase domain of fibroblast growth factor receptor in complex with inhibitors
Science
1997
, vol. 
276
 (pg. 
955
-
960
)
26
Itoh
H.
Hattori
Y.
Sakamoto
H.
Ishii
H.
Kishi
T.
Sasaki
H.
Yoshida
T.
Koono
M.
Sugimura
T.
Terada
M.
Preferential alternative splicing in cancer generates a K-sam messenger RNA with higher transforming activity
Cancer Res.
1994
, vol. 
54
 (pg. 
3237
-
3241
)
27
Eswarakumar
V. P.
Lax
I.
Schlessinger
J.
Cellular signaling by fibroblast growth factor receptors
Cytokine Growth Factor Rev.
2005
, vol. 
16
 (pg. 
139
-
149
)
28
Ibrahimi
O. A.
Zhang
F.
Eliseenkova
A. V.
Linhardt
R. J.
Mohammadi
M.
Proline to arginine mutations in FGF receptors 1 and 3 result in Pfeiffer and Muenke craniosynostosis syndromes through enhancement of FGF binding affinity
Hum. Mol. Genet.
2004
, vol. 
13
 (pg. 
69
-
78
)
29
Schüller
A. C.
Ahmed
Z.
Ladbury
J. E.
Extracellular point mutations in FGFR2 result in elevated ERK1/2 activation and perturbation of neuronal differentiation
Biochem. J.
2008
, vol. 
410
 (pg. 
205
-
211
)
30
Dailey
L.
Ambrosetti
D.
Mansukhani
A.
Basilico
C.
Mechanisms underlying differential responses to FGF signaling
Cytokine Growth Factor Rev.
2005
, vol. 
16
 (pg. 
233
-
247
)
31
Shukla
V.
Coumoul
X.
Wang
R. H.
Kim
H. S.
Deng
C. X.
RNA interference and inhibition of MEK-ERK signaling prevent abnormal skeletal phenotypes in a mouse model of craniosynostosis
Nat. Genet.
2007
, vol. 
39
 (pg. 
1145
-
1150
)
32
Wu
Y.
Chen
Z.
Ullrich
A.
EGFR and FGFR signaling through FRS2 is subject to negative feedback control by ERK1/2
Biol. Chem.
2003
, vol. 
384
 (pg. 
1215
-
1226
)
33
Treanor
B.
Lanigan
P. M.
Suhling
K.
Schreiber
T.
Munro
I.
Neil
M. A.
Phillips
D.
Davis
D. M.
French
P. M.
Imaging fluorescence lifetime heterogeneity applied to GFP-tagged MHC protein at an immunological synapse
J. Microsc.
2005
, vol. 
217
 (pg. 
36
-
43
)
34
Suhling
K.
French
P. M.
Phillips
D.
Time-resolved fluorescence microscopy
Photochem. Photobiol. Sci.
2005
, vol. 
4
 (pg. 
13
-
22
)
35
Ibrahimi
O. A.
Zhang
F.
Eliseenkova
A. V.
Itoh
N.
Linhardt
R. J.
Mohammadi
M.
Biochemical analysis of pathogenic ligand-dependent FGFR2 mutations suggests distinct pathophysiological mechanisms for craniofacial and limb abnormalities
Hum. Mol. Genet.
2004
, vol. 
13
 (pg. 
2313
-
2324
)
36
Mohammadi
M.
Olsen
S. K.
Ibrahimi
O. A.
Structural basis for fibroblast growth factor receptor activation
Cytokine Growth Factor Rev.
2005
, vol. 
16
 (pg. 
107
-
137
)
37
Hatch
N. E.
Hudson
M.
Seto
M. L.
Cunningham
M. L.
Bothwell
M.
Intracellular retention, degradation, and signaling of glycosylation-deficient FGFR2 and craniosynostosis syndrome-associated FGFR2C278F
J. Biol. Chem.
2006
, vol. 
281
 (pg. 
27292
-
27305
)
38
Le
R. C.
Wrana
J. L.
Clathrin- and non-clathrin-mediated endocytic regulation of cell signalling
Nat. Rev. Mol. Cell Biol.
2005
, vol. 
6
 (pg. 
112
-
126
)
39
Haglund
K.
Sigismund
S.
Polo
S.
Szymkiewicz
I.
Di Fiore
P. P.
Dikic
I.
Multiple monoubiquitination of RTKs is sufficient for their endocytosis and degradation
Nat. Cell Biol.
2003
, vol. 
5
 (pg. 
461
-
466
)
40
Mosesson
Y.
Shtiegman
K.
Katz
M.
Zwang
Y.
Vereb
G.
Szollosi
J.
Yarden
Y.
Endocytosis of receptor tyrosine kinases is driven by monoubiquitylation, not polyubiquitylation
J. Biol. Chem.
2003
, vol. 
278
 (pg. 
21323
-
21326
)
41
Peschard
P.
Park
M.
Escape from Cbl-mediated downregulation: a recurrent theme for oncogenic deregulation of receptor tyrosine kinases
Cancer Cell
2003
, vol. 
3
 (pg. 
519
-
523
)
42
Waterman
H.
Katz
M.
Rubin
C.
Shtiegman
K.
Lavi
S.
Elson
A.
Jovin
T.
Yarden
Y.
A mutant EGF-receptor defective in ubiquitylation and endocytosis unveils a role for Grb2 in negative signaling
EMBO J.
2002
, vol. 
21
 (pg. 
303
-
313
)